Shrink a DataFrame with smaller dtypes

Pandas data engineering interview problem. Difficulty: intermediate. Pattern: Data Quality. About 16 minutes. Part of the Pro drill bank.

Downcast wide numeric and object columns in a DataFrame to minimal safe memory types. Treat this as a production helper: match the contracted return shape, including empty and duplicate inputs.

Given a DataFrame df with columns id (int64), score (float64), and tier (object/string): Optimize the DataFrame's memory footprint by downcasting column data types to the smallest safe representations: Downcast all integer columns to the smallest suitable integer type (pd.to_numeric(..., downcast="integer")). Downcast all floating point columns to the smallest suitable float type (pd.to_numeric(..., downcast="float")). Convert text/object columns to 'category'. Return a summary DataFrame describing the optimized schema with columns: column: column name dtype: string representation of the downcasted column dtype (e.g. 'int8', 'float32', 'category') Sort by column ascending and reset the index. Assign the summary DataFrame to result.

Requirements

• Return column and dtype columns.
• Sort by column ascending.

Constraints

• Values are within valid numeric ranges.
• Object columns contain string categories.

Examples

Input: df (int64, float64, object) id | score | tier 1 | 12.5 | silver 2 | 23.0 | gold 3 | 45.5 | silver 4 | 60.0 | bronze 5 | 99.5 | gold Output: column | dtype id | int8 score | float32 tier | category Integer 1..5 fits in int8, float values downcast to float32, and repetitive string tier becomes category.

Topics: lakebench, pandas, downcast, category.

More interview problems · All interview problems · Learn data engineering